Here’s a step-by-step guide on how I made Jose, a Puerto Rican AI model.

What am I doing?

I’m creating a virtual AI fashion model named Jose, designed to look like a real Puerto Rican man. On the FASHN AI platform, this kind of virtual model is called a consistent model.

A consistent model is an AI-generated person who looks photorealistic and stays the same across all the images you generate, meaning he keeps the same face, features, and style every time you use him. Unlike random AI outputs that look different each time, a consistent model ensures that Jose always looks like “himself,” no matter the setting or pose.

Why am I doing this?

With a consistent model, I can create realistic photos of Jose in any situation I imagine, different poses, outfits, and backgrounds, while still looking like the same person each time. This is perfect for generating fashion shoots, lifestyle photos, or creative scenes featuring my AI model Jose.

How am I doing this?

First, I start with Imagen 3. Imagen is an image generation model from Google that generates high-quality, photorealistic images from text prompts. I used the Imagen playground on fal.ai to access the model. Imagen is also available through Google's Gemini here.

The "Face Seed"

I found that the following prompt worked best, for creating what I call the “face seed”. I generated a couple times, selected my favorite of the bunch, then moved on to the next step.

A passport photo of the full face of Jose, a fit Puerto Rican male wearing a black t-shirt, in his 20s, with well trimmed facial hair, with a white background. The photo itself is photorealistic, very high quality, as if it were taken by a 50MP Full-Frame Camera. Focus sharply on the face's details.

Imagen's output.

I called it the ‘face seed’ since it’s the starting point, a clear, realistic image of our soon to be consistent model Jose’s face, that I can build on to create many other images of him in different settings. You are free to modify this initial prompt in any way that you like. 

I am going to use this “face seed” to create multiple images of Jose in different settings with Midjourney, since FASHN requires 8-20 images to create a Consistent Model. I’ll shoot for 12 diverse images here.

Navigating Midjourney

Inside of Midjourney, I navigate to the create tab.

I enter this as the prompt:

full-body portrait of a 5'10" person with an athletic build, standing and looking straight forward, photorealistic style, city alleyway background, 50 MP full-frame camera detail --ar 3:4 --v 7 --style raw

This helps define the subject, pose, style, setting, and technical details to generate a realistic, high-res full-body portrait.

I also make sure to select “omni reference”, which will allow me to preserve the likeness of the ‘face seed’ that I created previously across my generations.

What the Midjourney UI looks like, with Omni Reference and the Face Seed selected.

And there we have it, our first set of 4 outputs to choose from.

Midjourney Output

I choose my favorite and move to the next prompt.

My favorite output of the 4 Midjourney generations.

For the next prompt, we repeat the process. We want to have a diverse set of photos, so for this one, we’ll put Jose in a swimsuit and vary the camera framing.

full-body photo of a 5'10" person with an athletic build, wearing swim shorts, standing at a 45-degree angle from the camera, photorealistic style, 50 MP full-frame camera detail --ar 3:4 --style raw --v 7

You can see my favorite of the 4 outputs here. In the output images that I pick, I want to avoid:

• hats, sunglasses, masks, or other people

• heavy filters or retouching

• near-duplicates, every image should be distinct

Jose in a Swimsuit

To save you time (and sanity) I repeated the process with 10 other images and recorded all of the prompts that I used below.

You're welcome to use these Midjourney prompts yourself with your own custom face seed.

Midjourney Prompts

Now that I have my photos, I upload the photos into the FASHN Studio’s Consistent Models page and click submit. It takes 3-5 minutes for the Consistent Model to be created.

Jose is Born

There we have it! Jose, our very own consistent model comes to life.

Now we can take Jose and start generating as many models of him to use for our virtual try-ons as we want in the Create Models tab of FASHN...

Prompts used in FASHN's Model Studio with Jose selected as Consistent Model.

It blows my mind that I can use AI to create an AI version of a person who doesn’t even exist.  

Now I can have Jose model anything that I want, anywhere I want.

I made our AI Model Jose try on various clothes with FASHN's Virtual Try On.

Back in 2022, fine‑tuning a DreamBooth model on my own GPU took almost nine hours; today FASHN creates Jose in a couple of minutes, and the whole thing costs $99 a month instead of the thousands you’d burn on a typical fashion‑shoot day. That’s a time‑to‑value drop from “all day and all budget” to "coffee break and pocket change".

Bring your own AI model to life with FASHN.